I see a lot of visualization around me now and I am extremely excited about it. Yet, are we making any real difference? I mean, are we having any real impact in people’s life other than telling them beautiful stories?

Yes I know, impact could be defined in a million different ways and it may be hard to capture. But why? Why I never stumble into an article or blog post showing, I don’t know, for instance, how visualization helped a group of doctors doing something remarkable with visualization?

Is it just because this stuff does not get reported or what?

Here are a few possible explanations:

• Explanation#1: Impactful visualization is hidden. Those people who are using visualization successfully, who have a real impact, are too busy to report their success.
• Explanation #2: Visualization is just a fragment of a much larger process. Visualization, when is not used as a communication/story telling tool is part of a much larger process, which includes many other steps and tools so simply success is not ascribed to visualization.
• Explanation #3: Visualization impact has yet to come. Maybe we just have to wait a bit longer and we’ll get all the success we want.

What do you think? Do you have other explanations? Is my question just too pretentious? Or did I just miss a ton of success stories and this post is totally nonsense?

P.S.1 On a side note: other areas of data analysis, especially automatic approaches like machine learning and data mining have plenty of stories to tell. Why? Food for thought …

P.S. 2 After writing this post I discovered my friend Andy Kirk has written a much longer post on this issue.

{ 4 comments }

Being a visualization researcher means a very large body of my work revolves around pushing the boundaries of visualization further. I do that by mostly developing innovative techniques but also trying to better understand how humans interact with this amazing tool we call visualization.

You might think I have at least a rough idea of what progress means in visualization then, but in fact I don’t. And I guess I am not alone: researchers are trained to dive into tiny details and speculate for ages. The purpose of this post is to explore bigger questions:

• What is progress in visualization?
• How do we make progress in visualization?
• And how do we measure it?

I ask that because honestly I don’t see a direction in what we are doing. We researchers are mostly focussed on developing yet another technique, practitioners on (understandably) satisfying their customers. But what is our ultimate goal? Here I propose s few ways we can look at progress in visualization.

Progress As Real-World Impact

First and foremost I propose progress in visualization is the extent to which we are able to help people do remarkably useful things with data. This is for me the gold standard, the holy grail. It is a broad and vague definition but it helps. When I say “remarkably useful” I mean: can we say visualization played a critical role in curing or preventing diseases? Reducing poverty? Solving or preventing economic crises? Make people richer or happier? Etc. Think about it, why not? Why do we do visualization if not for these purposes?

Despite some few isolated cases I don’t see this happening now. We should keep our eyes open and focus more on having an impact in the real world. Visualization has this potential, I am sure, and progress is made, I believe, when we help people do remarkable things. The VisWeek conference used to host a very nice session called Discovery Exhibition with the specific intent to showcase success stories. Unfortunately, (its hurts to admit it) I think it was quite a failure. I remember a similar frustrating post from Stephen Few some years ago: “True Stories about the Benefits of Data Visualization“. And I have yet to see persuasive answers to his call.

Progress As Knowledge Construction

I have to admit measuring progress exclusively in terms of impact and success stories might be a bit fuzzy, not very practical and ultimately a bit subjective. Another possibility is to define progress as the accumulation of knowledge that permits to build more effective visualization. But what do we need to know that we don’t know yet? Broadly speaking we need to know:

1. How humans work.
2. How to translate knowledge about humans into visualization design.

Are we doing that right now? Partly, in academic environments and a bit outside, but not enough in my opinion. It’s surprising to see how much more foundational work has been done in the past and how little today. We have a rough idea of how visual variables (position, length, color, size, etc.) work in isolation but very little understanding of how they interact in complex environments. We have alternative visualizations for the same kind of data and little understanding of how they influence information extraction (parallel coordinates vs. scatter plot matrix? node-link diagrams vs. matrices? maps or abstract representation? animation or small multiple?) And we have not even started scratching the surface of muddier issues like semantics, influence, persuasion, etc.

Progress as Technical Achievement

I don’t even know if I need to comment on this one, it’s pretty straightforward: technical achievement is the development of visualization and interaction techniques that solve unsolved technical problems or improve performance over existing solutions. Typically this takes the following form:

• New visualization or interaction design.
• Faster and/or more accurate algorithms.
• Increased scalability in terms of data size and dimensionality.
• Accommodation of new data formats and tasks.

I think it’s safe to say academic research is mostly focused on this. I am not sure whether technical achievement translates into real benefits in real-world applications but from time to time we have really useful stuff coming out. Edge bundling and horizon graphs are the first things that come into my mind. Are we making progress in this area? Yes. Would I like to see more? Yes and no … In a way sometimes I feel like we are spinning the wheel (please note that I include myself into this description and I am not immune to many many faults) so I’d like to see less spinning-the-wheel technical contributions and more useful stuff. But I also realize we cannot invent a new edge bundling every year. Progress happens with valleys and peaks.

Progress As Education and Adoption

Maybe this is the most neglected kind of progress, yet it very much lies at my heart. The last way to define progress in visualization I propose is the extent to which we are able to teach people how to judge and use visualization effectively and how many people will use visualization in their work. We need to reach more people (visualization at school?) but more importantly we need to teach proper visualization. We need courses, seminars, teaching material, web sites, and a whole army of evangelists. I am lucky enough to know quite a bunch of them but we need more.

I want to measure progress in a few years by counting how many people are able to criticize a chart. I also want to measure progress by assessing whether visualization will be part of the standard toolbox of scientists, business men and decision makers around the world.

Conclusion

This is what I had to say about progress. I know it’s not perfect, it’s just a draft. And now it’s your turn. How do you define progress in visualization? Are we making progress? How would you measure progress in visualization in, let’s say, 5 or 10 years from now?

And by the way, do you care about making progress? Why not? It is not necessary to be “a researcher” to make progress, you can make progress in a thousand ways. The only thing we need is to bring more focus. Or maybe we just have to let things happen and have some fun? I am looking forward to hearing from you guys. Thanks for reading.

—

On a side note: I have been out of the scenes with FILWD for a very long while. There are good reasons why that happened (I’ll tell you more about that later) but I want to assure you FILWD is not going to fade away. To the contrary, I have many plans on how to grow it further and offer a better service. If you are still there reading me after so much time well … thank you so much from the bottom of my heart! -Enrico

{ 9 comments }

I was reading the description of a new data visualization contest coming out today, the Nielsen Data Visualization Contest, and an apparently insignificant sentence caught my attention: “The challenge is to make data tell a story, conveying what’s most important effectively and efficiently.”

There is a lot of attention lately around using visualization to “tell a story” and I can understand why: visualization, when designed properly, has a tremendous effect on people. Not only it has the power to convey a clear message and to make complex concepts very easy to grasp, but it also has the power to persuade. I guess the main reason being that when a statement is backed up by data then people believe it is true(er).

I have nothing against using visualization to tell stories, to the contrary I am fascinated by this use of visualization and I think it’s very relevant. For instance, raising awareness about important facts or democratizing access to complex information are very noble intents of visual story telling, and I fully support them.

But, I don’t know, call me old-style, conservative, bigot: I am concerned by an excessive focus on story telling. It’s an itch I cannot scratch. And because I cannot express it in a closed form the only thing I can do is to make a list of concerns I have (hoping your comments will make it easier to dispel the fog).

There’s no story telling without data exploration. Creating a story with visualization doesn’t mean there is not role for data exploration in visualization in its making. People looking at the final product might think the power of visualization is exclusively in the effective presentation of the facts. But what people don’t see is the amount of exploratory work behind every story. I know as a matter of fact that many great visualization designers start with a thorough visual exploration of the data at hand using standard tools like Tableau or R. Without this preliminary phase it’s very hard to tell a compelling story and it is also very hard to come up with an enlightening visualization.

It’s the data that makes the story not the visualization. I always laugh a bit when people complain about David McCandless’ work. They say that their visualizations are not optimal and that he makes many “mistakes”. In a way I agree but why does he have such a big success then? I think the reason rests in his ability to select amazing stories to tell. The story is hidden in the data. Well, not even in the data, I guess everything starts in his mind, the rest just follows naturally. So, if we are passionate about visualization and dare about its proper use I believe story telling is (maybe) not the most challenging area to test it.

Many people need visualization to build our future not to tell a story. While I cannot resist a catchy well-crafted data visualization that tells a compelling story, I also know from my experience how desperately professionals of all kinds need visualization to just do their work best. I am talking about doctors, engineers, biologists, policy makers, etc. Part of our life, or of our future generations, might depend on them and we have the opportunity to help them help us. Don’t you think this use of visualization is a bit under represented on the web when compared to the whole set of story telling visualizations out there? For instance, why don’t we have contests to help these people with their data and have plenty of those asking to vaguely find a story to tell in this or that data set?

A story is not THE truth. I have no evidence for that but my feeling is that visualization can be used to more easily persuade people. By the mere fact of being built on top of data people might think it is truer than other kind of stories. Again, you can see that in McCandless’ work. Many of his pieces are evidently conceived to be provocative and touch hot topics. But I bet that for every provocative visualization out there there is the possibility to build a counter argument with another one. I might be proven wrong on that but I haven’t seen any evidence on the contrary so far.

Not all stories are worth telling. Since the power of a story resides in the data, it is not always possible to tell a compelling story. Regardless the beauty or inventiveness of your visualization if the data is dull you might not get a compelling story. And I have experienced it so many times that I am almost inclined to say that this is pretty much the standard for any given data set. You can see it in the recent Information is Beautiful Award: there are many cool and pretty entries, some that I really like from the design point of view, but is there anything really interesting there to see? Do we leave the stage enriched by new knowledge?

{ 12 comments }